An intervention that delayed onset of Alzheimer's disease (AD) by several years would yield huge public health benefits. Several studies suggest that non-steroidal anti-inflammatory drugs (NSAIDs) may produce such a delay. NSAIDs may also attenuate progressive age-related cognitive decline (ARCD) when this conditions represents a prodrome of AD. Both prevention strategies can be evaluated definitively only in randomized trials. Such trials can also examine attendant risks of long- term NSAID use in the moderate doses that appear to afford protection against AD and ARCD. Improved safety may be available with selective cyclooxygenase-2 (COX-2) inhibitors, but it is not clear that COX-2 inhibition offers the protective effect apparent with conventional NSAIDs. We therefore propose a parallel trial of the common NSAID ibuprofen and the selective COX-2 inhibitor celecoxib vs. placebo for prevention of AD and for attenuation of ARCD. The trial will involve four sites and enroll 2625 dementia-free subjects aged 72-88 with a history of Alzheimer-like dementia for a first degree relative. Therefore, a conspicuous decline in periodic cognitive screening test results will identify subjects with suspected incident dementia. We will evaluate these subjects clinically using structured, standardized methods of assessment and diagnosis. The proposed sample presumes 7 years of observation, with realistic estimates of attrition through mortality and other causes, and of treatment "drop-outs" and "drop-ins." It should provide 80% power (2-tailed a=0.05) to detect a 30% reduction in incidence among the treated groups. In this application we proposed the first 42-54 months of treatment with an interim analysis of efficacy after the last-enrolled subject has completed 30 months. At that point, the study will have 80% power to detect a 50% reduction in AD incidence with either agent, in which case the trial can be stopped. The trial should also be stropped if there is no apparent benefit of treatment, or if safety issues mandate. Otherwise, the intermit estimate of treatment effects will dictate the shape of a competing renewal application an additional 0.25 to 4 years of observation and a final analysis. As a secondary outcome, we will examine the trajectory of cognitive scores to assess ARCD.